-
Notifications
You must be signed in to change notification settings - Fork 51
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improve the overall performance #109
Comments
For reference, Autoware.Auto also did it similarly: |
Hi @xmfcx, thank you for your proposal!
We are currently implementing the improvements discussed in the working group 3 weeks ago autowarefoundation#4024, namely a performant/usable straight-to-disk recording solution and per-packet decoding (without aggregating a whole scan). PR will be out shortly. Also, thank you for pointing out |
Thank you @xmfcx, as we briefly discussed yesterday, I think we are mostly in agreement with some extra comments as below:
Looking forward to putting together a plan 👍 |
I'm currently working on step 1 of this: Packets are processed as they arrive, and not sent over DDS or intra-process comms but handled in one wrapper node altogether. It's still in an early stage but should drastically reduce Nebula latency. |
Related Issue:
Description
Currently, the driver has a lot of latency which can be avoided by some key changes.
Especially for large point clouds.
What can be done?
Proposed architecture
This way, we eliminate complexity, increase efficiency, eliminate cpu load spikes.
Support multiple point types
And if we want to support multiple point cloud types, we can maintain multiple raw PointCloud2 types and in the last step of aggregation, we can add them to these messages before publishing.
cc. @drwnz @mojomex @miursh @yukkysaito @kminoda
The text was updated successfully, but these errors were encountered: